DE102016222640A1 - Monitoring system, method, in particular for the detection of manufacturing errors, and use of a monitoring system - Google Patents

Abstract

A monitoring system for evaluating one or more properties of a product, in particular prior to its startup, comprising a data structure comprising archived first and second data, the archived first data comprising information from a product creation process of at least one product of one type, and wherein the archived second data is property data of the at least one product, and wherein the monitoring system comprises a computing system configured to correlate the archived first and second data such that based on current first data of a product of that type, prediction of second data of the product is possible.

Description

The present invention relates to a monitoring system, in particular for the evaluation of one or more properties of a product, a method, in particular for the detection of manufacturing defects, and a use of a monitoring system.

There are components which can not be completely tested for their functionality in a non-destructive way or where non-destructive test methods would not be economical. This is z. B. the case of crankshafts. Although other components can be tested non-destructively, they must be subjected to specific test cycles or test procedures. The problem here is the evaluation or interpretation of the data obtained in the test cycles. Often, the review is done manually, and the decision on whether a value is considered "okay" or "out of order" is usually based on the experience of the individual auditor and not on a traceable schema. The same applies if the component is already in operation. Here it is known to use sensors of any kind and the data obtained z. B. on the basis of a defined scheme, whereupon a statement about the state of the component can be made. Here, however, only one actual state can always be mapped. Neither is it possible to predict the future behavior of the component, nor can the causes of possible errors be identified or uncovered.

It is therefore an object of the present invention to provide a monitoring system, a method, in particular for the detection of manufacturing defects, and a use of a monitoring system which eliminate the aforementioned disadvantages and in particular allow a prediction of component behavior.

This object is achieved by a monitoring system according to claim 1, a method according to claim 10 and by a use according to claim 12. Further advantages and features will become apparent from the dependent claims and the description and the accompanying figures.

According to the invention, a monitoring system for assessing or predicting one or more properties of a product, in particular before it is put into operation, comprises a data structure comprising archived first and archived second data, the archived first data comprising information from a product development process of at least one product of one type, and wherein the archived second data comprises property data of the at least one product, and wherein the monitoring system comprises a computing system or an evaluation unit configured to relate the archived first and archived second data to each other based on current data or actual or just recorded / measured, first data of a product of this type, a prediction, prediction or prognosis of second data of the product is possible or possible.

The property data are in particular quality data and / or data from the operation of the corresponding product. The property data thus contains information about properties of the corresponding product, these properties being e.g. For example, information may include whether a product is in order or not, whether or not it has shown some damage, in other words, whether it is a B. failed or not.

There are no limits on the type of product. In preferred embodiments, in particular components or components for the automotive industry. This can be complex products such as engines, which include a variety of individual components act, but also to individual components such as crankcases, cylinder heads, pistons, crankshafts, etc.

The data structure expediently has one or more storage systems for storing the archived first and second data, for example in the form of data records. The storage preferably takes place in such a way that the correlating or associated first and second data can be assigned to one another, wherein the functionalities required for this purpose can be made available via data warehouses and / or data lakes. For example, a database is provided in order to map the data structure, with particular preference being given to electronic memory as possible data storage, which also includes cloud computing. It is thus expedient to ensure that the archived first data and the archived second data of one and the same product are assigned to one another. By now correlating the datasets of a plurality of products of the same type with each other, it is possible to recognize dependencies and conspicuousnesses, etc. It is thus possible to compare the information from the product development process with the information from the operation of a product or with the property data and derive insights therefrom. Advantageously, this information is used on the basis of currently detected, for example, measured first data of a product, during the product development process or on At the end of the product development process, to forecast whether a process step was successful or how the product will behave in the future, in other words, whether it is qualitatively in order. Conveniently, for. B. based on currently detected first data are detected that an engine, for example, after 7000 operating hours, suffered damage, such as a piston clamp. It is particularly advantageous in this context that a variety of different information can be used. The use of a wide variety of data during the product development process makes it possible to detect or recognize dependencies that would otherwise be unidentifiable. Thus, a product can go through a wide variety of process steps, with each process step possibly being classified as "OK" in its own right, but the combination of these process steps, which are in itself good, can lead to a defective product. The evaluation of the current first data can be carried out continuously during the entire product development process, whereby a rapid response and in particular a quick troubleshooting, for example in the form of a post-processing, is made possible. Possibly. Troubleshooting can also be to carry out downstream process steps differently, in other words "adapted".

According to one embodiment, the first data comprises production data, wherein the monitoring system is designed to capture and / or process the production data. Production data can be measured values of production equipment, assembly parameters such as bolting data, operating data of the respective equipment (such as power consumption), etc. For this purpose, the monitoring system is expediently equipped with corresponding sensors or sensor systems or has them. According to one embodiment, the monitoring system is also designed so that it can use the corresponding sensors and actuators, which is used in the tools and devices used in the product anyway anyway, or z. B. can read. These current first data can then be compared with the archived first data, which are stored in the data structure and which contain information on already tested and delivered engines. Based on the comparison can be conveniently determined whether z. B. an engine is faulty or wherein the cause of the fault is.

According to one embodiment, the first data comprises test data, in particular cold test data, wherein the monitoring system is designed to record or process the test data. In this case too, the monitoring system is expediently designed to access the sensors or actuators of corresponding test stands or test runs and to process these data. Alternatively, the monitoring system itself may also be designed such that it can detect such test data itself or has interfaces for the input and output of this data. Alternatively or additionally, it is also possible to combine a plurality of first data or data sets with one another, for example a plurality of test data records or also different data records such as test data records and production data, such as possible screwing data, etc.

According to one embodiment, the test data include z. B. Data from end-function tests such. B., in the case of an internal combustion engine, from a hot test or cold test, in particular so for example cold test data. The cold test is a test procedure in the assembly process of an internal combustion engine, in which it is checked whether the previous assembly steps have been carried out correctly. The engine is driven passively by an electric motor and not on its own. Appropriately, here structure-borne sound data are recorded and recorded. The characteristics of structure-borne noise data and cold test data etc. generate data profiles for specific engine types and causes of faults. The analysis of the test data and their characterization can be done by different methods. Structure-borne sound signals can be broken down into specific order spectra by means of a Fourier transformation, for example. As a result, individual sensor data can be arranged in different patterns. In addition, a speed-dependent analysis of the data can be carried out, with the help of which abnormalities can be mapped on a specific crankshaft angle. Both the raw data and the transformations thereof can be used for the analysis.

Of particular advantage is the non-destructive and rapid testing of components. Conveniently, the monitoring system z. B. be used to detect a possible grinding fire. The grinding burn is a thermally induced damage of ground workpiece edge zones and can occur if too much heat develops during the grinding of hardened surfaces. This creates tempering zones, new hardening zones and even cracks, with altered hardness and reduced compressive stress or even tensile stress. The grinding process can also be detected non-destructively via etching processes, but the known processes are relatively time-consuming. Advantageously, therefore, with the monitoring system currently detected during the grinding first data, for example in the form of structure-borne sound data, can be used to predict property data of a product, such as a camshaft or crankshaft. Stored models based on archived first and second data are used as a basis to predict whether the camshaft or the crankshaft z. B. grinding burn is present.

According to one embodiment, the first data also include rework data, test bench data, measured values of production systems, assembly parameters such as bolting data, operating data of the respective installation (such as power consumption), etc.

Conveniently, the computing system is designed to perform pattern recognition. The production data as well as the test data, for example, the cold test data, the archived first data, as well as the second data are z. B. subjected to a (mathematical) pattern recognition. Pattern recognition is to be understood as recognizing regularities, repetitions, similarities or laws in a set of data. Expediently, the collected archived first data and / or the second data can therefore be analyzed against this background, and the knowledge gained thereby or the knowledge gained thereby can be used to interpret the currently acquired first data. Thus, it can be predicted or predicted or predicted whether the product in question will fail in operation or in the field once, for which purpose information from the product development process, ie before the actual commissioning of the product, is used. With great advantage, therefore, a prognosis of the future product behavior or the future component properties is possible. In computer science, different approaches to pattern recognition are currently known, which can be divided into syntactic, statistical or structural. Conveniently, the monitoring system is configured to provide one or more such approaches.

Alternatively, the archived first and second data also z. B. evaluated by hand and z. B. be recorded statistically. For example, it is possible to create decision trees in order to correlate the archived first and second data with each other or to recognize patterns and derive knowledge. These findings can be rewritten into (mathematical) models, which allow for the precalculation of second data.

In one embodiment, the computing system is configured to create a model relationship between the archived first and / or the archived second data, wherein the model relationship is configured to process the current first data or at least portions thereof as input variables. The archived first and / or second data are expediently used to construct a model, for example a mathematical model, which processes the currently acquired first data as input variables and simulates / predicts future second data of this product on the basis of the currently acquired first data. The archived first and second data are thus used to train or adapt a mathematical model. In this case, classical polynomial models or neural networks, etc. can be used, wherein the model type is not limited to the aforementioned two examples.

As the models are based on archived data, among other things, they continue to grow and are automatically optimized. Expediently, the "relationship" of the data also takes place based on errors. If z. If, for example, enough data is available for a specific error type or the characteristic data (characteristic set of first data) has been recognized for this type of error, a suitable model can be created on this basis and used to identify, for example, vehicles that have the error Although not yet in operation, but this may have soon. These vehicles can then z. B. be recalled or it will, possibly in the context of the next workshop visit, carried out a recall.

According to one embodiment, the computing system is also designed to compare the current first data with each other or with the archived first data. For example, there is an archived first record that "belongs" to a product that has not failed in the field / farm. This data record can be used as a reference data record, with which the currently acquired first data are compared, whereby preferably limit values are determined by means of which it can be recognized whether the product will be free of errors. For example, information from pattern recognition can also be used for this purpose.

Expediently, the computing system is also designed such that it can identify a characteristic set of first data and / or also second data, wherein, for example, the aforementioned pattern recognition is also used for this purpose. In this respect, it is not necessary to take into account all information from the product development process or also from the operation, but to create a characteristic data set which is sufficient for a sufficiently reliable assessment of future component behavior.

If there are already existing tests in the production process or in the manufacturing processes used, z. B. collected data also in characteristic patterns divided and used for the component state determination or the precalculation of the second data. In this case, both patterns are expediently considered, which consist of individual first data such as process parameters and z. B. give second data, as well as data having a characteristic combination of a plurality of first data such as process parameters or second data.

According to one embodiment, the first and the second data are recorded in a time and / or process triggered manner. This means that the recorded data can be assigned exactly to a specific production process or test bench trial. This makes it possible, for example, certain production data or test data, clearly associate with, for example, a specific engine behavior or vice versa. As a result, production processes can be improved with pinpoint accuracy or risky production stages can be identified. In addition, the comprehensive data collection and the targeted pattern recognition / characterization of these data can be used to reduce the tests or limit them to efficient test contents. It is particularly advantageous that characteristic data sets can be determined. Conveniently, unfavorable component pairings can be detected in this way, for example, if two components are installed from the tolerance limit range and it may cause errors.

According to one embodiment, the product comprises at least one component or a plurality of components, wherein the first and second data are recorded component-triggered. This means that the data or the data records, be it first or second data, can be assigned exactly to the respective components. The archived first and second data are thus concretely referenced to certain component properties. In the case of corresponding abnormalities of components or products in the field, the corresponding process parameters and production data can be specifically checked for patterns that occur again and again, for example, in the case of defective components or specific types of defects. These patterns can then again form a basis for decision making for currently manufactured components.

Conveniently, this can be achieved in that a component test is carried out on the basis of the information collected during production. Advantageously, this eliminates additional testing procedures. In particular, the aforementioned review can be done non-destructive and it can be dispensed with further intermediate test cycles.

• - Bereitstellen von archivierten zweiten Daten, wobei die zweiten Daten Eigenschaftsdaten zumindest eines Produkts eines Typs umfassen;
• - Bereitstellen von archivierten ersten Daten, wobei die ersten Daten Informationen aus dem Produktentstehungsprozess des zumindest einen Produkts umfassen;
• - Erkennen bzw. Analysieren von Abhängigkeiten und Beziehungen zwischen den archivierten ersten und zweiten Daten;
• - Erfassen von aktuellen ersten Daten während eines Produktentstehungsprozesses eines Produkts des Typs und Vorausberechnung von zweiten Daten auf Basis der aktuellen ersten Daten und den Abhängigkeiten und Beziehungen bzw. unter Berücksichtigung der Abhängigkeiten und/oder Beziehungen.

The invention is also directed to a method, in particular for the detection of manufacturing errors, comprising the steps:

• Providing archived second data, the second data comprising property data of at least one product of one type;
• - providing archived first data, the first data comprising information from the product creation process of the at least one product;
• - detecting or analyzing dependencies and relationships between the archived first and second data;
• Acquiring current first data during a product creation process of a product of the type and pre-computing second data based on the current first data and the dependencies and relationships, or considering dependencies and / or relationships.

For the method, the advantages and features mentioned in connection with the monitoring system apply analogously and correspondingly or vice versa.

• - wobei nach Inbetriebnahme des Produktes aktuelle zweite Daten als Eingangsgrößen für die Prognose mitverwendet werden.

Conveniently, the method further comprises the step:

• - After commissioning of the product current second data are also used as input variables for the forecast.

Expediently, the monitoring system therefore also continues to run during the operation or (further) use of the product. The stored patterns, models, comparison methods, etc. are thus further adapted or "learn" on. As part of a workshop visit, it is then possible, for example, to read out the monitoring system and make a forecast about the future product or component behavior. Conveniently, it is z. B. possible to recognize that a particular component will soon fail, although the current fault memory, such as the engine, shows no error.

Conveniently, the precalculated second data are also stored and stored in the data structure, whereby z. For example, it is documented that a component with the label "OK" has left the factory.

It should be mentioned at this point that the archived second data can be read out, for example, within the framework of various workshop visits from the control units of the vehicles. Alternatively or additionally, the archived second data also comes from customer surveys, Workshop surveys, etc. In order to perform the above comparison, for example in the context of a workshop visit, it may, for. B. be provided to store the archived first data and the currently detected first data of the corresponding or respective product in a corresponding storage medium of the vehicle so that the on-board electronics can access this information to provide the functionality of the monitoring system.

The invention also relates to the use of a monitoring system according to the invention for detecting defects, in particular manufacturing defects, in components or products. The use of the monitoring system is particularly preferred in a control unit of a motor vehicle.

All of the aforementioned features and advantages relating to the monitoring system and the method apply analogously and correspondingly to the use or vice versa and to one another.

Further advantages and features will become apparent from the following description of preferred embodiments of the monitoring system or the method or the use with reference to the accompanying figures. Different features can be combined with each other within the scope of the invention.

• 1: ein Schema eines Überwachungssystems;
• 2: ein weiteres Schema eines Überwachungssystems bzw. des entsprechenden Verfahrens zur Verdeutlichung von dessen Arbeitsweise.

Show it:

• 1 a scheme of a surveillance system;
• 2 : Another scheme of a monitoring system or the corresponding method to illustrate its operation.

1 shows a monitoring system comprising a data structure 10 which archived first data or data records 11 and archived second data or records 12 a product of a certain type. The archived second data 12 For example, customer surveys, workshop surveys, read-out control units, data from test vehicles, etc. are the first data 11 include information from the product development process, such as test data, test bench data, metrics from manufacturing equipment, etc. The monitoring system further includes a computing system 14 , The computing system 14 captures current first data 21 which have been or will be recorded as part of a current manufacturing or production process of a product of this type. The computing system 14 is designed, the currently recorded first data 21 as well as the collected or archived first data 11 and the collected or archived second data 12 to account for each other, to relate, to correlate, to model, etc. Thus, it is possible to calculate precalculated second data 22 which in turn is in the data structure 10 can be stored. In other words, therefore, a prognosis on the future component behavior is possible or it is a detection of manufacturing errors possible, which allows, in particular non-destructive, to check whether a component or product is in order or not. The basis for this is the currently recorded first data 21 or parts thereof. The currently recorded first data 21 are in turn stored in the data structure and thus archived second data 11 ,

2 essentially shows that 1 known monitoring system. With the reference number Z is a kind of timeline or a timeline shown. The reference number Z ' represents, for example, the time of commissioning of a product, such as a vehicle, comprising an internal combustion engine and / or an electric motor, currently recorded first data 21 are analyzed and collected / archived first and second data 11 and 12 a data structure 10 from a computing system 14 related to each other. This results in precalculated second data 22 created. These can now provide information as to whether the product, for example the engine, is in order or not or in particular whether it will fail or not. The first dates 21 come, as it were, on a virtual test bench, which in the 2 with the reference number 50 outlined. Are the precalculated second data 22 okay, can a release 52 be granted. Are the precalculated second data 22 not ok, will be a troubleshooting 51 carried out. The corresponding product or component is then subjected to, for example, a rework. The virtual test bench 50 allows not only a review of the current state of the component, for example, "okay" or "not in order" but also a prediction or a possibility of predicting that the engine is still "alright" even after 10,000 hours. With the reference number Z " is a time of a workshop visit sketched. Here are current second data 23 read and processed by the monitoring system, which is shown for example in the control unit of the corresponding vehicle. Although, for example, the error memory or present in the vehicle fault memory indicate no error, it is thus possible to predict or predict whether the vehicle z. B. would soon fail, whether, for example, a piston clamp is to be expected, etc.

After all, data of the products striking the plant / field can be ideally used. Analyzes of causes of errors can be carried out, which lead to time and cost savings. This makes it possible, for example, to send the appropriate product, such as the engine, into the appropriate repair. Testing steps in manufacturing processes and warranty costs can be reduced. Image damage is avoided. Automatically, production processes can be improved. A non-destructive and, in particular, very fast testing of components is also made possible. This makes it possible, above all, to carry out a one hundred percent exam, not just a random exam. If necessary, previously required test methods can be restricted and critical production steps can be identified. The great advantage of the monitoring system, the method or the use of the monitoring system is in particular that the system is learned and not static.

LIST OF REFERENCE NUMBERS

10

data structure

11

archived first data / records

12

archived second data / records

14

computing system

21

current first data

22

precalculated second data

23

current second data

50

virtual test bench

51

troubleshooting

52

release

Z

time beam

Z '

Time, startup

Z "

Time, during operation, (workshop visit)

Claims (12)

Monitoring system for evaluating one or more properties of a product, in particular before it is put into operation, comprising a data structure (10) comprising archived first (11) and second data (12), wherein the archived first data (11) contains information from a product development process of at least one Products of one type, and wherein the archived second data (12) includes property data of the at least one product, and wherein the monitoring system comprises a computing system (14) configured to interleave the archived first and second data (11, 12) / relate to each other based on current first data (21) of a product of this type prediction of second data (22) of the product is possible. Monitoring system after Claim 1 wherein the first data (11, 21) comprises production data, and wherein the monitoring system is configured to acquire the production data. Monitoring system after Claim 1 or 2 wherein the first data (11, 21) comprises test data, in particular cold test data, and wherein the monitoring system is designed to detect the test data. Monitoring system after Claim 3 , wherein the test data comprise structure-borne sound signals. A surveillance system according to any one of the preceding claims, wherein the computing system (14) is adapted to perform pattern recognition. The monitoring system of any one of the preceding claims, wherein the computing system (14) is adapted to create a model relationship between the archived first (11) and second data (12), and wherein the model relationship is arranged, the current first data (21) as inputs to process. A monitoring system according to any one of the preceding claims, wherein the computing system (14) is adapted to compare the current first data (11) with the archived first data to (21). Monitoring system according to one of the preceding claims, wherein the first (11, 21) and the second data (12) are recorded in a time and / or process triggered manner. A monitoring system according to any one of the preceding claims, wherein the product comprises at least one component, and wherein the first (11, 21) and second data (12) are component-triggered recorded. A method, in particular for the detection of production errors, comprising the steps: - providing archived second data (12), the second data (12) comprising property data of at least one product of one type; - providing archived first data (11), the first data (11) comprising information from the product development process of the at least one product; - detecting or analyzing dependencies and relationships between the archived first (11) and second data (12); - acquiring current first data (21) during a product creation process of a product of the type and precomputing second data (22) based on the current first data (21) and the dependencies and / or relationships. Method according to Claim 10 comprising the step: - after the product has been put into operation, current second data (23) are used as input variables for the prognosis. Use of a monitoring system according to one of Claims 1 - 9 for the detection of errors, in particular manufacturing errors, in components.

Images